Global AI Scale-up Interconnect Market By Technology, Component, Domain Size, End User - Market Size, Industry Dynamics, Opportunity Analysis and Forecast For 2026-2035

The AI scale-up interconnect market is undergoing an explosive structural transformation as artificial intelligence infrastructure shifts toward increasingly dense, tightly integrated computing architectures. The market is estimated at approximately USD 6.0 billion in 2025 and is projected to reach USD 60 billion by 2035, representing a compound annual growth rate (CAGR) of 26.0% over the 2026-2035 forecast period.

A fundamental driver of this transformation is the industry's movement away from conventional node-to-node networking architectures toward scale-up computing environments. Traditional scale-out infrastructure distributes workloads across multiple servers or nodes and relies on networking technologies to connect these independent systems. While this approach remains important for expanding overall computing capacity, the requirements of modern AI workloads are creating a greater need for tightly coupled systems in which large numbers of accelerators can communicate at extremely high bandwidth and very low latency.

Noteworthy Market Developments

Broadcom, Marvell, Astera Labs, Credo, and MaxLinear occupy important positions across the AI scale-up interconnect value chain. Their respective strengths span switch silicon, co-packaged optics, optical interconnects, DSPs, PCIe connectivity, active electrical cables, and high-speed signaling technologies.

The competitive landscape among these companies reflects the increasingly fragmented and specialized nature of the AI scale-up interconnect market. Broadcom is strongly positioned around switch silicon and CPO, Marvell combines optical connectivity, DSP expertise, and support for open standards, Astera Labs specializes in PCIe and advanced data-center connectivity, Credo focuses on AEC-based high-speed electrical connections, and MaxLinear contributes critical PAM4 DSP capabilities. Rather than competing across exactly the same product categories, these companies occupy complementary layers of the broader interconnect ecosystem.

As AI models become larger and accelerator clusters become increasingly dense, the need for high-bandwidth, low-latency, power-efficient connectivity will continue to intensify. These companies are therefore positioned to benefit from the structural expansion of AI infrastructure, while their differing technological specializations will continue to shape the development of next-generation scale-up interconnect architectures.

Core Growth Driver

A primary catalyst for demand in the AI scale-up interconnect market in 2026 is the growing physical limitation of traditional data-transfer architectures. As frontier artificial intelligence models expand toward and beyond multi-trillion-parameter scales, the performance constraints of AI systems are increasingly determined not only by the raw processing capability of GPUs but also by the speed and efficiency with which those processors can communicate. The enormous computational requirements of advanced models are forcing data-center operators to deploy increasingly large accelerator clusters, making high-bandwidth, low-latency connectivity a fundamental requirement for maintaining overall system performance.

Emerging Opportunity Trends

The transition toward silicon photonics, co-packaged optics (CPO), and optical circuit switching (OCS) is emerging as a significant opportunity for growth in the AI scale-up interconnect market. The rapid escalation of AI accelerator performance is creating an equally rapid increase in data-transfer requirements, placing substantial pressure on conventional electrical interconnect technologies. As data rates approach approximately 200 Gbps per lane and continue to increase, copper-based cabling faces growing challenges related to attenuation, signal integrity, power consumption, and transmission distance. These limitations are encouraging data-center operators and infrastructure manufacturers to accelerate investment in optical technologies capable of supporting the bandwidth requirements of next-generation AI clusters.

Barriers to Optimization

Fragmentation and the continued proliferation of proprietary interconnect protocols may hamper the growth of the AI scale-up interconnect market by creating uncertainty around interoperability, platform compatibility, and long-term infrastructure investment. As AI clusters become increasingly sophisticated, enterprises and hyperscalers require communication technologies that can connect large numbers of accelerators, memory resources, and other computing components efficiently. However, the coexistence of proprietary technologies and open industry standards can result in competing architectural approaches, making it more difficult for customers to establish universally compatible infrastructure.

Detailed Market Segmentation

By technology, NVLink and NVLink Fusion represent the leading technology category in the AI scale-up interconnect market, supported by the rapid expansion of generative AI workloads and the increasing computational demands of advanced foundation models. As AI systems progress toward models with extremely large parameter counts, conventional point-to-point connectivity and standard networking lanes face growing pressure to move data between increasingly numerous accelerators. This is encouraging a structural shift toward high-bandwidth, low-latency interconnect architectures designed specifically to allow large numbers of GPUs and AI accelerators to function as a tightly integrated computing environment.

By component, optical modules represent the leading segment of the AI scale-up interconnect market, driven by the fundamental limitations of conventional copper-based connectivity as data rates continue to increase. The rapid expansion of AI accelerator clusters is pushing interconnect speeds toward and beyond 800G, creating increasingly demanding requirements for bandwidth, signal integrity, reach, power efficiency, and thermal management. Although copper remains effective for shorter and lower-speed connections, its electrical transmission characteristics become increasingly challenging at very high data rates and longer distances.

By domain size, the Above 72 category leads the AI scale-up interconnect market in 2025, reflecting the rapidly increasing scale and complexity of modern artificial intelligence infrastructure. The exponential growth of foundation models, particularly large language, multimodal, and generative AI systems, is driving organizations toward increasingly expansive accelerator clusters. As models become larger and training datasets become more complex, individual GPUs are no longer sufficient to deliver the required computational capacity.

By end user, hyperscalers represent the dominant segment of the market, supported by their exceptional capital expenditure capacity, extensive infrastructure footprints, and aggressive investment in proprietary AI factories. The scale of computing infrastructure required for modern artificial intelligence workloads has created a market environment in which only the largest technology and cloud providers can consistently commit the enormous financial resources necessary to build and operate advanced AI clusters.

Segment Breakdown

By Technology

• NVLink /NVLink Fusion
• UALink
• Ethernet-Based Scale-Up
• Proprietary Fabrics

By Component

• Switch Silicon
• Link Controllers & IP
• Retimers & Redrivers
• Copper Cabling & Backplanes
• Optical Modules

By Domain Size

• Up to 8 Accelerators
• 8-72 Accelerators
• Above 72 (Rack/Pod Scale)

By End User

• AI Chip Vendors
• Hyperscalers
• Server OEMs
• Neocloud Providers

By Region

• North America
• The U.S.
• Canada
• Mexico
• Europe
• Western Europe
• The UK
• Germany
• France
• Italy
• Spain
• Rest of Western Europe
• Eastern Europe
• Poland
• Russia
• Rest of Eastern Europe
• Asia Pacific
• China
• India
• Japan
• Australia & New Zealand
• South Korea
• ASEAN
• Rest of Asia Pacific
• Middle East & Africa (MEA)
• Saudi Arabia
• South Africa
• UAE
• Rest of MEA
• South America
• Argentina
• Brazil
• Rest of South America

Geography Breakdown

• North America captured the largest share of the market in 2025, supported by exceptionally high levels of capital expenditure from Tier-1 hyperscale technology companies and cloud service providers. The region has emerged as a central hub for the development and deployment of advanced artificial intelligence infrastructure, with hyperscalers committing substantial resources to data centers, accelerator clusters, high-performance computing platforms, and next-generation networking systems.
• The United States serves as the primary engine of this regional dominance because it hosts many of the world's most influential semiconductor designers, AI technology companies, cloud infrastructure providers, and hyperscale computing operators. The concentration of these companies within the U.S. creates a powerful technology ecosystem in which semiconductor innovation, AI model development, data-center construction, and cloud deployment reinforce one another.
• The region's mature cloud-computing ecosystem provides another structural advantage. Major cloud providers can deploy AI infrastructure at enormous scale and make advanced computing resources available to enterprises through cloud services. This allows organizations that cannot independently construct large AI factories to access accelerator capacity through hosted infrastructure. As cloud providers expand their AI capacity, demand for the underlying high-bandwidth networking infrastructure increases correspondingly.

Leading Market Participants

• NVIDIA
• Broadcom
• Marvell Technology
• Astera Labs
• Credo Technology
• Alchip
• Cadence
• Synopsys
• AMD
• Enfabrica
• Amphenol
• TE Connectivity
• Molex
• Rambus
• Micas Networks
• Other Prominent Players